AgentTeamLandAI agent teams, installed like packages.

A package manager and registry for curated teams of AI agents — install a stack, inherit another team's work, ship.

📦

Teams as packages

A team bundles specialized agents, skills, and rules for a kind of work — full-stack apps, data pipelines, mobile products. Install once, symlinked into your project.

🧬

Inheritance done right

Single-parent extends with override and excludes. Unlimited depth, circular detection, caret version constraints — npm-grade semantics, no diamond problem.

⚡

One static binary

atl is a ~7 MB Go binary. Install via Homebrew, Scoop, winget, or curl — zero runtime dependencies.

🧪

Battle-tested

The reference team software-project-team ships 13 agents covering .NET, Flutter, React, and a production-grade Docker stack. End-to-end verified.

🔍

Public registry

Discover teams by name. The registry is a single JSON file, PR-driven, schema-validated in CI.

🛠️

Open and scriptable

Every piece is MIT-licensed. team.json is a public schema. Build your own team and submit it.

See it in action

In 30 seconds

bash

# macOS / Linux
brew install agentteamland/tap/atl

# Windows
scoop bucket add agentteamland https://github.com/agentteamland/scoop-bucket
scoop install atl

# Then, in any project:
atl install software-project-team

That's a full 13-agent stack — API, web, mobile, database, infra, review — wired into your project's .claude/ directory, ready for Claude Code.

What's next?

What is atl? — the big idea in five minutes.
Quickstart — first team installed in under a minute.
Team authoring — publish your own team.